F-fluoride PET/CT for assessing bone involvement in prostate and breast cancers

18
F-fluoride PET/CT for assessing bone involvement
in prostate and breast cancers
Nadia Withofs
a
, Benjamin Grayet
b
, Tino Tancredi
b
, Andre
´e Rorive
c
,
Christine Mella
e
, Fabrice Giacomelli
e
, Fre
´de ´ric Mievis
e
, Joe
¨l Aerts
e
,
David Waltregny
d
, Guy Jerusalem
c
and Roland Hustinx
a
Objective To evaluate the accuracy of
18
F-fluoride
PET/computed tomography (CT) to detect bone
metastases (BMs) in a breast and prostate cancer
population, using magnetic resonance imaging (MRI)
or thin-slice CT as a gold standard.
Methods We have prospectively included 34 patients
with breast (N= 24) or prostate cancer (N= 10) at high
risk of BMs. Whole-body PET/CT (low-dose CT) and bone
scintigraphy (BS) with single photon emission CT were
obtained for all 34 patients and the results compared with
a radiological gold standard.
Results Out of the 386 foci detected by PET/CT, 219
(56.7%) could be verified by CT or MRI. Eighty-six
additional foci were detected by BS (n= 46) or seen only
by CT (n= 9), MRI (n= 23), or both CT and MRI (n= 8).
The total number of verified lesions was therefore 274
(58.1%), including 119 (43.4%) benign and 155 (56.6%) BM.
The sensitivity, specificity, and accuracy of
18
F-fluoride
PET/CT were 76, 84.2, and 80%, respectively. For BS,
they were 44.8, 79.2, and 60%, respectively. Sensitivity
significantly decreased for the lytic lesions. The accuracy
of PET/CT was significantly superior to BS for pelvic
and lumbar lesions. PET/CT provided a correct diagnosis
(M + /M0) in 32 of 33 patients (one false positive)
compared with 28 of 33 with BS (four false positive,
one false positive).
Conclusion
18
F-fluoride PET/CT is significantly
more accurate than BS for detecting BMs from
breast and prostate cancers. Nucl Med Commun
32:168–176 c2011 Wolters Kluwer Health | Lippincott
Williams & Wilkins.
Nuclear Medicine Communications 2011, 32:168–176
Keywords: bone metastases, breast cancer,
18
F-fluoride, PET/computed
tomography, prostate cancer
a
Division of Nuclear Medicine,
b
Department of Medical Imaging,
c
Division of
Medical Oncology,
d
Division of Urology, CHU Lie
`ge and
e
Cyclotron Research
Center, University of Lie
`ge, Lie `ge, Belgium
Correspondence to Dr Roland Hustinx, MD, PhD, Division of Nuclear Medicine,
Centre Hospitalier Universitaire de Lie
`ge, Sart Tilman B35, Lie `ge I 4000, Belgium
Tel: + 32 43667199; fax: + 32 43668257;
e-mail: [email protected]
Received 11 July 2010 Revised 4 September 2010
Accepted 5 October 2010
Introduction
Prostate and breast cancers are the most common
malignancies worldwide, with a high incidence of bone
metastases (BMs). Early detection and accurate assessment
of bone involvement is needed to optimize treatment and
therefore reduce or delay skeletal-related events. Currently,
whole-body bone scintigraphy (BS) is only recommended
in selected patients at high risk of BM and not for routine
breast or prostate cancer surveillance [1,2]. Computed
tomography (CT) and magnetic resonance imaging (MRI)
are sensitive and specific modalities for the diagnosis of
BM, but are limited to an anatomical region [3,4]. MRI is
superior to CT for detecting early intramedullary lesions
[5]. Newer whole-body techniques of MRI acquisition are
promising but are not fully validated for routine clinical use
[5]. Compared with conventional imaging, 2-deoxy-2-
[
18
F]fluoro-D-glucose (FDG) positron emission tomogra-
phy (PET) seems to exhibit higher specificity and accuracy
to detect BM in breast cancer [6]. The results of studies
comparing FDG-PET to BS are conflicting but the latter
is generally considered the best option for assessing
the entire skeleton [7]. In contrast, the contribution of
FDG PET in prostate cancer is limited. Indeed, FDG
uptake can be low especially in well-differentiated tumors
limiting the sensitivity of FDG-PET for the staging or
follow-up of prostate cancer [8]. Alternative PET tracers
are under investigation for prostate cancer imaging.
Recent studies evaluating
18
F-choline PET/CT showed
promising result for the detection of early prostate cancer
BM [9,10].
Blau et al. [11] first imaged the skeleton with
18
F-fluoride in
the early 1960s.
18
F-fluoride is easily produced in high spe-
cific activity. Its pharmacokinetic properties are superior to
those of technetium-99m (
99m
Tc) methylene diphospho-
nate (MDP) used for BS resulting in a higher bone uptake
and faster blood clearance [12,13]. As current PET systems
combine high sensitivity and spatial resolution with re-
duced acquisition time,
18
F-fluoride is being actively re-
evaluated. Compared with BS, all studies showed higher
sensitivity of
18
F-fluoride PET to detect BM, even for lytic
lesions [14–18]. Only few data suggest a higher accuracy of
PET/CT with
18
F-fluoride to detect BM compared with
conventional techniques [19,20].
Original article
0143-3636 c2011 Wolters Kluwer Health | Lippincott Williams & Wilkins DOI: 10.1097/MNM.0b013e3283412ef5
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

The aim of this study was to evaluate, in a population of
breast and prostate cancers, the accuracy of
18
F-fluoride
PET/CT and
99m
Tc-MDP BS for detecting BMs, using
full diagnostic CT or MRI as a gold standard (GS).
Methods
Patients
We prospectively included 34 patients with breast
(n= 24) or prostate (n= 10) cancer at a high risk of BM
(mean age ± standard deviation: 60.2 ± 12.3 years). The
protocol was approved by the ethics committee of our
institution and all patients gave informed consent to
participate in the study. We included patients with
asymptomatic suspect bone lesion on BS (15 female
patients) or elevated cancer antigen CA 15-3 greater than
or equal to 45 UI/ml (N= 3) or prostate-specific antigen
greater than 20 ng/ml (N= 9) or clinical suspicion of
bone involvement (one female; one male). Finally, five
breast cancer patients with known BM were included for
restaging during treatment (tamoxifen citrate, N=2;
aromatase inhibitor, N= 2; weekly paclitaxel, N= 1).
Twenty-three patients (18 females; five males) were on
antihormone therapy before PET/CT imaging and 16
patients with breast cancer received chemotherapy in
the course of the disease. Out of these 16 patients who
received chemotherapy, one received paclitaxel 15 days
before PET/CT, one received neoadjuvant chemotherapy
(FEC 100: epirubicin, 5-fluorouracil, and cyclophospha-
mide) 20 days before PET/CT and 14 received che-
motherapy but not recently (median time: 659.5 months;
minimum: 20 months; maximum: 1235. 2 months). Ten
patients received biphosphonates (oral biphosphonate: five
female patients; intravenous zoledronic acid: five female
patients administered to all at least 1 month before PET/
CT imaging).
18
F-fluoride PET/computed tomography
18
F-fluoride was prepared by proton irradiation of water
enriched in oxygen-18 with an IBA 18/9 cyclotron (Ion
Beam Applications, Louvain-La-Neuve, Belgium) using
targets made of niobium or silver and windows made of
Havar with a usual beam of 30 min. The radioactive
solution was then transferred to a GE MX synthesizer
(General Electric, Loncin, Belgium) modified to accept a
tuned kit that contained all the raw materials needed.
After adsorption on an anion exchange resin and washing
with water for injection, the
18
F-fluoride was eluted with
a sterile physiological solution. The solution was then
dispensed with sterilizing filtration according to Euro-
pean cyclic GMP. Quality control of the sodium
18
F-
fluoride solution was done as described in the European
Pharmacopoeia (01-2008 : 2100). All patients received
300 MBq of
18
F-fluoride through an indwelling catheter.
We asked the patients to walk and drink 1 l of water
during the uptake time of 45–60 min. Thirty-one PET/
CT studies were undertaken using a Gemini Dual system
(Philips, Cleveland, Ohio, USA), which combines a GSO
crystal-based PET and a dual-slice CT scanner. Three
were acquired using a Gemini TF (Philips, 16-slice CT).
A whole-body low-dose CT acquisition (5 mm slice
thickness; tube voltage: 120 kV and tube current–time
product: 80 mAs) was followed by a PET scan. PET
acquisition time in the Gemini Dual system was 2 min
per bed position (pbp) from the skull to the upper thighs
and then 1 min pbp to the toes. In the Gemini TF
system, it was 1.5 min pbp for the axial skeleton and
1 min for lower limbs. Data were corrected for decay,
scatter, random, attenuation and were reconstructed
using an iterative three-dimensional row-action maximum
likelihood algorithm; CT data were used for attenuation
correction.
99m
Tc-methylene diphosphonate bone scintigraphy
Whole-body planar BS (PBS) was performed 3–4 h after
an injected dose of 1000 MBq
99m
Tc-MDP using a
double-headed g-camera (e.cam; Siemens, Erlangen,
Germany), fitted with low-energy high-resolution collima-
tors, 1024 256 matrix, acquisition speed of 15 cm/min.
Thirty-four single field-of-view (FOV) single-photon emis-
sion computed tomography (SPECT) scans were done
using the same machine (same collimators, 128/128 matrix,
step-and-shoot method with a 1801rotation, non-circular
orbit, without zoom, 32 steps of 20 s). Twenty were
centered on the thoracic region and 14 on the lumbar spine
and pelvis. Data were reconstructed on a Mirage work-
station (Segami Corporation, Columbia, USA) using an
iterative algorithm (ReSpect).
Computed tomography and MRI
Thin-slice CT images (limited to a region of high
18
F-
fluoride uptake) were acquired using standard clinical
parameters, directly after the PET/low-dose CT on the
Gemini Dual system (N= 39; maximum three regions per
patient) or in the Department of Radiology with a multi-
slice spiral CT (N= 1; Siemens Somaton Sensation 16).
Twenty-two MRI were taken using three different
systems (Harmony 1 Tesla, Symphony 1.5 Tesla, and
Trio 3.0; Siemens). Acquisition sequences were similar
for each studied region: T1-weighted sequence [repeti-
tion time (TR), 550 ms; echo time (TE), 15 ms; flip angle
(FA), 9031], short TI inversion recovery sequences (TR,
4000 ms; TE, 63 ms; FA, 1801) and T1-weighted fat
saturation sequences (TR, 550 ms; TE, 15 ms; FA 901)
after an intravenous injection of 20 ml gadolinium.
Image analysis
Two nuclear medicine physicians interpreted whole-body
PET/CT images (corrected and not corrected for attenu-
ation). They were not aware of the indication of the
PET/CT (symptoms, biological markers, or suspicious
lesion on BS). All foci of
18
F-fluoride increased uptake
were recorded. A radiologist analyzed the low-dose CT
18
F-fluoride PET/CT and bone metastases Withofs et al. 169
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

images in parallel. A consensus was then established to
classify each lesion as malignant (score 4), most likely
malignant (score 3), equivocal (score 2), most likely
benign (score 1), or benign (score 0). The radiologist
characterized the lytic, sclerotic, or mixed patterns of
malignant lesions. The choice of the GS was based on the
localization of lesions. When the lesions were localized on
the skull, thoracic cage, or long bone, a centered diag-
nostic CT scan was performed directly after the PET/
CT acquisition in the hybrid system. For the vertebral
column or the pelvic lesions, an MRI scan was planned in
the department of radiology. We did not consider CT or
MRI confirmation for joint lesions or lesions located on
hands, feet, or knee except when any doubt on malig-
nancy existed. All diagnostic CT scans were taken in the
hybrid system directly after the PET/CT except for one
acquired 13 days later. The median time interval between
PET/CT imaging and MRI acquired in the radiology
department was 6 days (range: 0–28 days). A second radio-
logist independently visualized thin-slice CT and MRI
and categorized each lesion as benign or malignant. BS
imaging was performed before or after PET/CT with a
median time interval of 14 days (range: 1–35 days). An
experimented nuclear medicine physician blindly re-
viewed BS images, the lesion classes being the same as for
the PET/CT.
Data analysis
In a lesion-based analysis, we confronted the diagnosis
obtained with
18
F-fluoride PET/CT and BS with the con-
clusions of the diagnostic CT or MRI scans. Second, we
separated the lesions with regard to their localization:
skull and cervical spine, dorsal or lumbar spine, pelvis,
thoracic cage, and long bones. In the third analysis, we
considered the sclerotic, lytic, or mixed characteristics of
metastases.
In a patient-based analysis, we considered the final
diagnosis (metastatic or not) of the PET/CT and BS
compared with the GS for each patient.
Statistical analysis
Sensitivity, specificity, positive predictive value, negative
predictive value, and accuracy were calculated at the
lesion and patient levels for each PET/CT and BS
technique.
We compared the different imaging techniques using the
area under the curves obtained with receiver operating
characteristic analysis. A Pvalue inferior to 0.05 was
considered statistically significant.
The ability of each technique to detect BM was also
compared using the McNemar statistical hypothesis tests
with a Pvalue of less than 0.05 defined as being stati-
stically significant.
Results
Lesion-based analysis
Three hundred and eighty-six foci of
18
F-fluoride
increased uptake were detected, and 219 (56.7%) of
them were verified by either MRI (N= 121) or CT
scanning (N= 76) or both (N= 22).
18
F-fluoride PET/CT
described 136 BM, 62 benign lesions, and 21 equivocal
lesions. There was a total of 274 (58.1%) verified lesions
including additional foci detected on BS and verified by a GS
(N= 15) and lesions seen on CT and/or MRI only (N=40).
According to diagnostic CT or MRI, 119 lesions (43.4%) were
benign and 155 (56.6%) were BM. Out of 53 breast cancer
metastases specified with thin-slice CT, 24 were sclerotic
(45.3%), 12 lytic (22.6%), and 17 were mixed (32.1%).
Considering inconclusive lesions as benign, the sensitiv-
ity of
18
F-fluoride PET/CT and BS was 76 and 44.8%,
respectively. Table 1 summarizes the results in the two
populations considering an inconclusive lesion as benign
or malignant. Specificity for each technique was 84.2 and
79.2%, respectively. Receiver operating characteristic
Table 1 Sensitivity, specificity, PPV, NPV, and accuracy of
18
F-fluoride (
18
F-NaF) PET/CT,
99m
Tc-MDP BS with SPECT
Population
Sensitivity
(%)
Specificity
(%) PPV (%) NPV (%) Accuracy
18
F-NaF PET/CT
Breast 73.9/81.0 79.3/68.3 86.1/81.6 63.7/67.5 0.76/0.76
Prostate 100.0/100.0 94.7/89.5 85.7/75.0 100.0/100.0 0.96/0.92
Total 76.0/82.5 84.2/75.0 86.0/80.9 73.2/76.9 0.80/0.79
BS
Breast 43.0/43.0 76.8/70.7 76.3/71.8 43.8/41.7 0.55/0.53
Prostate 66.7/66.7 84.2/81.6 57.1/53.3 88.9/88.6 0.80/0.78
Total 44.8/44.8 79.2/74.2 73.4/69.0 52.8/51.1 0.60/0.58
Inconclusive lesions considered benign/malignant.
BS, bone scintigraphy; CT, computed tomography; MDP, methylene diphos-
phonate; NPV, negative predictive value; PPV, positive predictive value; SPECT,
single photon emission computed tomography.
Table 2 Sensitivity, specificity and accuracy of
18
F-fluoride PET/CT
and
99m
Tc-MDP BS in the various regions of the skeleton
Region Sensitivity (%) Specificity (%) Accuracy
Skull and cervical spine
PET/CT 66.7 86.7/80.0 0.83/0.78
BS 33.3 93.3/86.7 0.83/0.78
Dorsal spine
PET/CT 71.4/74.3 85.0/80.0 0.76/0.76
BS 51.4 90.0/85.0 0.65/0.64
Lumbar spine
PET/CT 73.1/84.6 82.6/82.6 0.78/0.84
BS 34.6 69.6/60.9 0.51/0.47
Pelvis
PET/CT 74.1/82.8 68.4/68.4 0.73/0.79
BS 31.0 73.7 0.42
Thorax
PET/CT 95.5 87.5/81.3 0.92/0.89
BS 81.8 61.9 0.72
Long bones
PET/CT 63.6/72.7 92.3/61.5 0.84/0.65
BS 45.5 90.5 0.75
Equivocal lesions considered benign/malignant (when it differed).
BS, bone scintigraphy; CT, computed tomography; MDP, methylene diphos-
phonate.
170 Nuclear Medicine Communications 2011, Vol 32 No 3
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

analysis showed that the ability of
18
F-fluoride PET/CT to
detect BM was significantly superior to BS (P<0.0001).
When analyzing PBS and SPECT, it seemed that the
contribution of SPECT to PBS was not significant
(P= 0.3124). Table 2 summarizes sensitivity, specificity,
positive predictive value, negative predictive value,
and accuracy of
18
F-fluoride PET/CT and PBS in different
locations of the skeleton. Compared with BS, the
sensitivity of PET/CT was higher in all locations. The
accuracy of PET/CT was significantly superior to BS
for pelvic and lumbar spine locations (P< 0.05).
Figure 1 illustrates an osteoblastic BM located on a
lumbar vertebra identified with PET/CT and overlooked
with BS.
The sensitivity of PET/CT and PBS was related to the
heterogeneity of presentation of BM. Table 3 shows the
sensitivity of each technique for detecting lytic, sclerotic, or
mixed BM.
Patient-based analysis
For this analysis we excluded one patient with prostate
cancer. He was claustrophobic and refused MRI; a thin-slice
CT scan verified only one of 43 detected lesions. Thirty-
two patients out of the 33 (97%) were correctly diagnosed
(M +/M0: 20/12) with
18
F-fluoride PET/CT. PET/CT
erroneously characterized only one patient as being meta-
static. All patients with BM were identified with PET/CT.
Fig. 1
(a)
(b)
(c)
(d)
(e)
An osteoblastic bone metastasis (breast cancer) located on the fifth lumbar vertebra was identified with
18
F-fluoride PET/computed tomography (CT)
(transverse sections; a: PET; b: low-dose CT; c: fused PET/CT) and overlooked with bone scintigraphy (d). MRI (e: T1-weighted fat saturation
sequences after intravenous injection of gadolinium, sagittal view) confirmed the malignant character of the lesion (arrow).
18
F-fluoride PET/CT and bone metastases Withofs et al. 171
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

The extent of the disease was correctly estimated in seven
patients, overestimated in three, and underestimated in
10 patients. BS identified 19 of the 20 metastatic patients
(95%); the extent of the disease was correctly estimated
in three patients, overestimated in three, and under-
estimated in 13 patients. BS was false positive (FP) for
four patients and false negative (FN) for one patient. The
final diagnosis by PET/CT and BS was concordant for 29
(87.9%) patients (19 true positive; nine true negative, and
one FP) and discordant for four (12.1%) patients.
18
F-
fluoride PET/CT correctly estimated these four patients
(one true positive; three true negative); BS missed one
metastatic patient and falsely diagnosed metastases for
three patients. Three patients showed equivocal lesions
with PET/CT, none of them had shown BM on the basis of
MRI or thin-slice CT scanning. Two patients showed
equivocal lesions with BS and none of them had shown
BM on the basis of MRI or thin-slice CT scanning. At the
level of the patient, the difference between PET/CT and
BS was not statistically significant (P> 0.05).
Discussion
Bone involvement is frequent in breast and prostate
cancers, particularly in cases of recurrence. Because of its
ability to detect BM several months before plain radio-
graphy,
99m
Tc-MDP BS is the technique of choice for
screening. Nevertheless, the limited specificity of BS
often requires confirmation with CT scanning or MRI.
SPECT increases the sensitivity of the technique and the
new SPECT/CT hybrid system improves the ability to
distinguish BM from a benign lesion [21]. The availability
of SPECT/CT is growing but is not widespread yet and
planar imaging with a single FOV SPECT remains the
mainstay in many institutions.
Since the 1990s, few studies have evaluated the ability of
18
F-fluoride PET to detect BM. In a prospective study
including five patients with multiple skeletal metastases
from breast cancer, Petren-Mallmin et al. [14] reported a
high tracer uptake in both sclerotic and lytic breast cancer
BMs. The investigators suggested a dual tracer approach
combining FDG and
18
F-fluoride with encouraging results
[22,23]. All the initial studies comparing BS with
18
F-
fluoride PET showed a higher accuracy of PET in
diagnosing bone involvement [16–18]. More recently,
Kruger et al. [24] compared the diagnostic accuracy of
FDG-PET/CT with
18
F-fluoride PET in a population of
68 patients with nonsmall cell lung carcinoma. All BMs
Table 3 Results of
18
F-fluoride PET/CT and
99m
Tc-MDP BS in the
three types of BM in breast cancer
Metastases TP FN Sensitivity (%)
Osteolytic
PET/CT 7 5 58.3
BS 4 8 33.3
Osteoblastic
PET/CT 20/22 2/0 90.9
BS 16 6 72.7
Mixed
PET/CT 14 3 82.4
BS 8 9 47.1
Inconclusive lesions are considered benign/malignant (when it differed).
BM, bone metastases; BS, bone scintigraphy; CT, computed tomography; FN,
false negative; MDP, methylene diphosphonate; TP, true positive.
Fig. 2
(a)
(d)
(e)
(b) (c)
(f)
A bone metastasis (prostate cancer) was detected with
18
F-fluoride PET/computed tomography (CT) in the left hip-bone (transverse sections; a:
PET; b: low-dose CT; c: fused PET/CT images). There was no lesion individualized with
99m
Tc-methylene diphosphonate planar bone scintigraphy
(d), single photon emission computed tomography (e) only showed a slight asymmetry that was overlooked. MRI (f: T1-weighted fat saturation
sequences after intravenous injection of gadolinium, coronal view) confirmed the malignant character of the lesion (arrow).
172 Nuclear Medicine Communications 2011, Vol 32 No 3
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.